Mistral Launches Voxtral TTS: Open-Weight Voice AI
Mistral AI releases Voxtral TTS, a 4-billion parameter open-weight streaming text-to-speech model supporting low-latency multilingual voice generation with implications for synthetic media.
TikTok's AI Ad Disclosure Policy Is Failing Users
TikTok requires advertisers to label AI-generated content in ads, but major brands like Samsung are skirting the rules. The policy gap raises urgent questions about synthetic media transparency at scale.
Evaluating Deepfake Detection Tools for Enterprise
As deepfake threats escalate, enterprises need rigorous frameworks to evaluate detection tools. Key criteria include accuracy benchmarks, latency, modality coverage, and integration capabilities.
UK Report: Deepfake Detection Tech Firms Face Key Hurdles
A new UK government report identifies significant challenges facing companies developing deepfake detection technology, including evolving generative AI capabilities, market fragmentation, and trust gaps.
Deepfake X-Rays Fool Radiologists and AI Detectors
A new study reveals that AI-generated fake X-rays can deceive both experienced radiologists and automated detection systems, raising urgent concerns about synthetic media threats in healthcare.
Synthetic Identity Fraud May Hit $58B as Deepfakes Grow
Synthetic identity fraud is projected to reach $58.3 billion, with deepfake technology accelerating the creation of convincing fake identities that bypass traditional verification systems.
Survey Maps AI Security Landscape for Foundation Models
A new comprehensive survey provides a unified framework for understanding AI security threats across foundation models, covering adversarial attacks, deepfake generation, synthetic media detection, and content authenticity challenges.
Detecting Prompt Attacks via LLM Judges and Model Ensembles
New research proposes combining LLM-as-a-Judge with Mixture-of-Models to detect prompt injection attacks, a growing threat to generative AI systems including video and image generators.
Exons-Detect: A New Method to Spot AI Text via Hidden States
New research introduces Exons-Detect, which identifies AI-generated text by analyzing hidden-state discrepancies in exonic tokens—key linguistic markers that differ between human and machine writing.
Deepfake Cyberattacks Exploit Trust, Bypass Defenses
Info-Tech Research Group finds deepfake-powered cyberattacks increasingly exploit human trust to circumvent traditional security systems, urging organizations to rethink defense strategies.
EU Backs Nudify App Ban, Delays Key AI Act Rules
The European Union has endorsed banning AI-powered nudification apps while pushing back deadlines for its landmark AI Act. The move directly targets non-consensual synthetic intimate imagery.
Tencent Open-Sources Covo-Audio: 7B Speech Model
Tencent AI releases Covo-Audio, a 7-billion parameter open-source speech language model with a real-time inference pipeline for audio conversations and reasoning, advancing synthetic voice capabilities.